[Boards: 3 / a / aco / adv / an / asp / b / biz / c / cgl / ck / cm / co / d / diy / e / fa / fit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mu / n / news / o / out / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / t / tg / toy / trash / trv / tv / u / v / vg / vp / vr / w / wg / wsg / wsr / x / y ] [Search | Home]
4Archive logo
How hard would it be to build an exo suit...
Images are sometimes not shown due to bandwidth/network limitations. Refreshing the page usually helps.

You are currently reading a thread in /diy/ - Do It yourself

Thread replies: 47
Thread images: 10
File: image.jpg (66 KB, 537x371) Image search: [iqdb] [SauceNao] [Google]
66 KB, 537x371
How hard would it be to build an exo suit like HAL?
It seems to me keeping the legs balanced would be the hardest part
not hard if you have the skill and resources

very hard if you dont
>easy if you know how, hard if you don't
Hmm i'm making a robotic hand that mirror my other hands moves and i can tell that it's not that hard. Also i make it with lego ev3 so i don't have the best tools.
Are you using any online instruction or just winging it
Very hard. The difficulty us detecting your movement. How does the robot know if your are trying to move our just resting your hand.

A separate hand that mimics your motion is easy, but I've that conforms to your movement and assists is super difficult.
Do on my own
I have gyro sensors on my other hand so i only need to get the rotation it gives to a variable and the motor rotation in another variable. The motor rotation is the difference between the variables
Oh the last "rotation" word supposed to be "rotate"
Just make sure Your exoskeleton does not move to a position Your limbs are incapable of reaching
Lol no. Good luck doing fine manipulation tasks with that.

The HAL is going to be difficult to make. Good luck making the actuators. Getting the harmonic drives necessary is not going to be easy or cheap.

If you want to replicate the HAL exactly then you also need to figure out how the heck they decode electrical signals sent to motors.

Alternatively, you could use torque controlled actuators and amplify the torque applied by the wearer. See the BLEEX for example of a system that does this.
about $100,000 hard
Oh boy i told you i just put gyro sensors to my arm and rotate the motors like my arm rotate. I already made this so don't tell me i can't. Also the calculation is very easy.
I am not questioning whether you did it, I am questioning the efficacy. You may be able to move the robot by moving your arm, but can you do anything useful with it?
I can grab and hold things like a plastic bottle. It's not supposed to have any special use. I just had time and these "bricks" so i built it. Btw i'm a highschool student.
Extremely difficult given the system it uses. IIRC, the first American to use one described a potentiometer on the belt that changed force application/ratio.

Turned all the way up, just trying to lift his leg a little made his leg shoot up.

The biggest problem with exoskeletons is the power supply, since most operate off batteries. Some hobbyists have made smaller ones using a diving tank and "air muscles".


This is also a huge concern. If you don't use electrodes on your body, you're going to need a way to calibrate it so you can't overextend your limbs and rip them off.

If I had the money to try and develop an exo or larger mech, I'd go for a small propane engine mounted on the back running a hydraulic pump.

Hydraulic motors for the limbs, using variable vane impellers so you can adjust the ratio of speed vs torque on the fly.
Anyone know what the ratheon sarcos exoskeleton used for sensors? They used servo motors and were made back around 2007. With the advances in MCUs hobbyists should be able to build something similar now.

Looking at it I'm pretty confident I could build something similar, some modern MCUs could probably command every servo motor in the entire skeleton. But the control mechanism is a complete mystery, is it using pressure sensors in every joint, or is it just sensors in critical points with some sort of IK system moving the others.

are shit. Sure they are torquey, but they are less efficient, less responsive, and require more maintenance than electric actuators.

Using variable vane impellers for the limbs is a bit odd, especially if you really mean variable vane impeller and not variable displacement hydraulic motor. For one it makes things really complicated and less compact. Impellers do not work well at low speeds and are really more of high speed low torque devices.

The propane engine driving everything has some problems. For one, you waste power if you aren't doing much of anything. When you move a bunch of limbs at once the engine can struggle to keep up, which is one of the problems Boston Dynamic's ATLAS robots had in the DARPA challenge.

But hey, that's just like your opinion man.

Now in my opinion the best way to build a DIY exoskeleton would be to use water-cooled brushless motors driven by capacitors. This is what SCHAFT used to dominate the first DRC. You can buy water cooled brushless motors that have similar specs to what SCHAFT used from china for cheap. For capacitors, EV regenerative braking capacitors might work and are cheap. Only difficult part would be getting the damn harmonic drives.

Here's more on the actuators SCHAFT used.
File: sarcos-mechanism.png (124 KB, 432x235) Image search: [iqdb] [SauceNao] [Google]
124 KB, 432x235
>>They used servo motors
nope they used hydraulics.

>>some modern MCUs could probably command every servo motor in the entire skeleton.
computation is cheap, sadly actuators are not(well yet at least)

>>is it using pressure sensors in every joint, or is it just sensors in critical points with some sort of IK system moving the others.
We don't really know. We do know the operator holds onto something that measures force on it(pic related). So it probably does something like the IK system you described, probably moving to decrease force applied by the operator. Although they probably use pressure sensors too.

Sarcos has a humanoid robot that has force sensors on every joint, so it would not be surprising that their exoskeleton does the same:
Hardest thing in the tech industry is thinking up cool acronyms for your inventions
Everywhere says the XOS 2 uses servo motors with an IC powerpack. Maybe the XOS-1 was hydraulic. The hydraulic ones always look slow and unresponsive.

Actually I found a way to make cheap actuators, there's a bunch of direct drive 3 phase BLDC motors that can be turned into high powered servos with an encoder, I already have a motor and servo controller with a canbus I'm in the process of modifying it for high voltage for another project but it should be able to let you have 500w- 1kw servos per joint.
According to the below it uses high pressure hydraulics:

>> direct drive BLDC
Direct drive is inefficient, to get reasonable amounts of torque you need big bulky motors. It has been tried, pic related. Cogging can also be a problem for fine control.

Sounds pretty neat though. I am curious as to how much torque these brushless motors of yours have. Especially if they're those big gimbal motors.

You might also want to look into the control approaches for brushless gimbals. Most don't use encoders, but are still able to get position control.
Does anyone make compact high speed hydraulic servo motors? I have never seen any that move like the XOS actuators, unless they made their own customer hydraulics.

What I'm getting from a lot of these posts is it would be better easier to build a robot and control it remotely then try and build a power suit.
You don't need high speed, you need high torque. Humans don't spin their limbs around at hundreds of rpm.

As far as hydraulic servos go there is this: http://rnd.knrsys.com/english/view.html?id_no=13&PHPSESSID=423b75a92f994e743d9c09b44dcc8a70

It ain't cheap, neither are the control valves, or the other associated hydraulic equipment

well if you are trying to build a robot with the full capabilities of a human, that's gonna be hard no matter what.

Anything that requires more force than hobby servos becomes pretty hard for amateurs to build.

Now if you mean wheeled robot you could remotely drive around with your shit in it, yeah you could do that
I work for a drywall company

Id like something that could hold a sheet against the wall / ceiling while I move it into place and screw it in
File: sheet-rock-jack.jpg (13 KB, 350x350) Image search: [iqdb] [SauceNao] [Google]
13 KB, 350x350
a drywall jack isn't good enough?

>especially if you really mean variable vane impeller and not variable displacement hydraulic motor

Just looked it up, this is what I was imagining. I've worked with boat hydraulics before but I don't know a whole lot about them, didn't even know there was such a thing.

The only reason I said propane was so it would be safer to used in enclosed spaces, and it's a bit easier to just fill up a fuel tank than to wait for batteries to recharge.

Looking up ATLAS now, didn't know it had been tried already. Thanks for that.

I didn't realize that capacitors had advanced enough to be used for something like this. Pretty freaking cool.

I guess I just hate the thought of being dependent on batteries. It seems like every day I read about a new battery technology that promises greater density, service life, etc. but nothing ever changes.

Probably is the best bet though. What I do know is that hydraulic tanks can get pretty fucking hot and I can't imagine lugging that around in a suit all day.
If i cant have a robot suit i want a robot atleast
>I read about new battery technology, but nothing ever changes.
File: 0Xzxk9g.jpg (636 KB, 2496x1872) Image search: [iqdb] [SauceNao] [Google]
636 KB, 2496x1872
I wonder how much torque you need for a exoskeleton with performance like the XOS.

You can get these hoverboard and self balancing vehicle BLDC motors for next to nothing, even RC hobby engines cost 4x as much as them.

With no cooling and their stock limits I think they're around 20nm at stall. But with water cooling combined with the high voltage and high current motor control I'm building you might be able to get 100nm? I need to find out how roughly many nm the largest muscles in human limbs generate.
yikes! The wiring on that hoverboard motor looks nasty.

>>water cooling
The reason SCHAFT was able to cool their motors was because all the windings on said motors were in direct contact with the motor case. (pic related)

Those hoverboard motors would be very difficult to water cool. I don't even know how you'd get tubing in there.

>> I need to find out how roughly many nm the largest muscles in human limbs generate.
for an exoskeleton it is more important to have actuators with a large torque density, that is the ratio or torque to weight. If your actuators are putting all their torque just into lifting their own weight they are sort of useless.

Let's compare the torque density of a standard servo to this hoverboard motor.

for servo
From here:
weight is 39 grams-f
torque is 6.5 kgf *cm
so torque to weight is 1.67 m

For the hoverboard motor
from here: http://www.amazon.com/Replacement-Hoverboard-Balancing-Electric-Unicycle/dp/B017LG2D32
torque is 100 N*m
using weight of 3kgf

torque to weight is 0.68 m

So not as great as our servo. However, we can get a pretty good torque to weight ratio just by gearing it down 1:5 or so, no modifications to the motor.

Are you sure hoverboard motors really have that much stall torque?
well I can lift myself with one arm. so I'd assume at least 1000nm considering that requires overcoming gravity and my arm is a little shorter than a meter and i weigh 75kg

The haptics involved are extremely difficult.

Unless you spend excessive amounts of time, you will find your exo exhausting or chafing. A robot arm is easy enough because there's no delay or error between parts. Combining a human arm with a robot arm will result in deviation which will result in wear issues.
>Are you sure hoverboard motors really have that much stall torque?
This 3kg BLDC has 60nm but it's designed for high RPM not low speed torque.

The funny thing about the Chinese direct drive hub motors is they mainly use them because they're too cheap to spend money on reduction gears. So they build them with maximum torque in mind, that's why it has a huge stator and copper sticking out everywhere. All they care about is having a stator with a high saturation limit and cramming as much copper as possible in.

The other thing is with cooling, the wires can actually take quite high temperatures, you just need to keep the heat away from the magnets which fail at around 90C

Switched reluctance motors might also be interesting for exoskeletons, they perform like BLDC but have no permanent magnets and will take as much current you can send without melting copper wires while stalled.
File: Motor_Render.png (296 KB, 804x608) Image search: [iqdb] [SauceNao] [Google]
296 KB, 804x608
MIT designed an "optimal actuator"...

It's a custom housing around a BLDC motor (380W with 2.96N*m stall torque) with a planetary gear (5.8:1). Damn it, someone needs to build that!

Seems like a hoverboard motor could work, though.
You can get geared bike motors from china for $50.
All you would have to do is change the gear ratio, add an encoder and put it in a more suitable housing.
The motor the cheetah used has a peak torque of 10 Nm. Though they may have upgraded to a motor with 30 Nm.

But yes, the tentative hoverboard motor specs look pretty good. I am not sure I trust those chinese motors. Cogging might also be a problem.

But hey if you wanna buy one and test it out that would be awesome

Page 67 of this details how to measure cogging torque in a simple manner:

The link above this one also has details of the custom motor drivers they used
I have a few hoverboard motors but getting my driver PCB produced is difficult, everywhere I try outsources production to china and can't deliver anything until march due to chinese new year.
Been having that problem too. The price difference is still worth it though.
The worst thing is I actually tried 3 local places, none of them make the PCBs here despite advertising it and just resell chinese made.

I'm going to get a quote from a japanese company.
you're making your own driver boards? You putting a microcontroller on them? Can you do regenerative braking?
File: pcb.png (50 KB, 532x810) Image search: [iqdb] [SauceNao] [Google]
50 KB, 532x810
I'm making a high powered version of this.

The software is already really good, sensored and sensorless FoC, automatic configuration for almost any BLDC, supports encoders, has canbus integrated and many other useful things.

The weak part is the hardware, it was built around a TI8302 mosfet driver which has many other functions built in, it's limited to 60v and prone to be destroyed as it's powered straight from the high voltage motor input. It's also very compact but it ends up being a problem for many applications.

So I have built a board that shares much of the same layout around the MCU but completely replaced the mosfet driver and half bridge section, in theory it should be able to scale to 500 volts, and large motors without any software modifications.

Hopefully if it goes well I can sell a range of boards using the software platform, ranging from RC, drones and robots to electric motorcycles and cars.
HAL3 uses gyro+EMG sensors

You could mount some surface electrodes over your leg muscles. Differental emg signal is easy to amplify. You can use some instrumental amplifier ICs, like INA128. Amplified signal requires filtering to pass 20Hz to about 200 Hz (or more, but less than 500Hz).if you're going to use microcontroler boards with 0-5V adc you need to rectify signal first. Google precision full-wave rectifier design from texas instruments. To make signal scalable it's good to smooth it with lowpass filter (my best try was at 0,7-1Hz. You can easily set duty cycle of pwm with this shape of signal. Every user has different muscle signal chracteristics, so you need to calibrate you device for every muscle i believe. You have gyros plus emg signals, so i think it's posible to write neural algorythm for controling some motors with gyro feedback. I'd consider mechanical design first. You could make simple computrer model to simulate your skeleton control with yor emg plus gyroscope signals (for example export solidworks assemblies to simscape ). Some ai (based on simple neural network) could learn correct motion response. I'd design exoskeleton control thxat way.
Isn't there a competition where people build powered suits like these and show off against each other? I'm pretty sure I heard about that on tv a few years ago.
That was years ago.
HAL also uses force sensors in the feet. I don't think it uses gyro sensors, do you have a source for that?
They might have some information on building a suit somewhere. Maybe a blog or forum. We get threads like this from time to time but wouldn't it be neat if some /diy/nasaurs actually went and built a suit? I don't know what it would be used for but it would certainly be cool.
Thread replies: 47
Thread images: 10
Thread DB ID: 508278

[Boards: 3 / a / aco / adv / an / asp / b / biz / c / cgl / ck / cm / co / d / diy / e / fa / fit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mu / n / news / o / out / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / t / tg / toy / trash / trv / tv / u / v / vg / vp / vr / w / wg / wsg / wsr / x / y] [Search | Home]

[Boards: 3 / a / aco / adv / an / asp / b / biz / c / cgl / ck / cm / co / d / diy / e / fa / fit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mu / n / news / o / out / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / t / tg / toy / trash / trv / tv / u / v / vg / vp / vr / w / wg / wsg / wsr / x / y] [Search | Home]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.
This is a 4chan archive - all of the shown content originated from that site. This means that 4Archive shows their content, archived. If you need information for a Poster - contact them.
If a post contains personal/copyrighted/illegal content, then use the post's [Report] link! If a post is not removed within 24h contact me at wtabusse@gmail.com with the post's information.